Printed Arabic optical character recognition using support vector machine

Abstract

Arabic Optical Character Recognition (AOCR) is the science of conversion Arabic text image documents of type printed or handwritten into editable text. OCR role is to help or replace humans in computerizing paperwork to accelerate, improve and reduce cost as well as time and effort. It also provides the ability to edit, store more compactly and search documents electronically. It is not a recent research field; it had started about 40 years ago. This paper proposes an Arabic OCR system that has five different stage. In segmentation stage, a new segmentation approach based on profile projection techniques are proposed. In the feature extraction stage, a combination of 56 statistical features are extracted, and for classification stage, we use Support Vector Machine classifier. The results show that the Character Recognition Rate achieved average of 99.08% for isolated Arabic characters, and 95.03 % for printed Arabic text.